Sensors
- TanDEM-X: X-band bistatic SAR pair producing a global DEM at 12 m posting (WorldDEM product) with absolute vertical accuracy around 2 m. The fine resolution makes it the primary source for cliff-face slope extraction and aspect derivation at range scale.
- Copernicus DEM GLO-30: Derived from TanDEM-X acquisitions and freely available at 30 m posting. Adequate for mountain-range-scale ruggedness mapping and initial candidate-cliff identification; loses fidelity on narrow cliff bands narrower than roughly 60 m.
- Sentinel-2 MSI: 13-band multispectral at 10 m (visible and NIR) and 20 m (red-edge, SWIR), 5-day revisit at mid-latitudes. Used to derive bare-rock spectral indices (NDVI suppression, SWIR reflectance) and to distinguish vegetated ledges from exposed rock faces.
- SPOT 6/7 or Pléiades: 1.5 m (SPOT 6/7) or 0.5 m (Pléiades) panchromatic resolution with on-demand tasking. The only freely accessible-orbit sensors capable of resolving individual ledge width and nest-platform geometry at site-confirmation scale.
Why cliff geometry is the bottleneck, not species behaviour
Cliff-nesting raptors, including Gyps vultures, Aquila eagles and Falco peregrines, are not randomly distributed across mountain terrain. They select for exposed rock faces above a minimum area threshold, with aspect preferences that vary by species and latitude, and with proximity to updraught sources that reduce the energetic cost of takeoff. Ground survey of these sites is expensive, dangerous, and impossible at the spatial scale needed for population-level conservation planning. A mountain range covering several thousand square kilometres may contain hundreds of candidate escarpments, of which perhaps a few dozen are occupied in any given season.
The practical question for a conservation programme is not where the birds are today, but where the physical conditions permit nesting at all. That is a geometry and spectral problem, and it is largely solvable from orbit, with one important caveat: the ledge itself, the actual nest platform, is typically 0.5 to 3 m wide. No freely available elevation product resolves that. The satellite can find the cliff; it cannot confirm the ledge without commercial tasking.
What a floating roof gives away: DEM-derived cliff metrics
The Copernicus DEM GLO-30 at 30 m posting, and the TanDEM-X WorldDEM at 12 m, both support slope-angle classification at the scale of escarpment bands. Cells with slope angles exceeding 55 to 60 degrees are reliably cliff or near-vertical rock; anything below 45 degrees is unlikely to support an open nest. Aspect is extracted from the same surface and is directly interpretable: south-facing cliffs in the northern hemisphere receive more solar radiation, which matters for egg incubation thermoregulation in early-season nesters such as Bearded Vultures (Gypaetus barbatus).
Terrain Ruggedness Index (TRI), computed as the mean absolute difference between a central cell and its eight neighbours, identifies the most topographically complex zones within a study area. High TRI values in combination with high slope and bare-rock spectral signal produce a composite suitability score. The method is well-established in the peer-reviewed literature on raptor habitat modelling and requires no field calibration to generate a ranked candidate list, though field calibration substantially improves predictive accuracy.
One honest limit: TanDEM-X penetrates dry snow and sparse vegetation to some degree in X-band, which can introduce small vertical biases in high-altitude terrain. In practice, for bare-rock cliff extraction, this is rarely a material problem, but analysts should validate DEM-derived slope values against any available field GPS transects in vegetated transition zones.
Reading bare rock from Sentinel-2 bands
Sentinel-2's 13 spectral bands allow bare-rock discrimination from vegetation and soil with reasonable confidence. The Normalised Difference Vegetation Index (NDVI) suppresses rock surfaces to values typically below 0.1, while vegetated ledges sit above 0.3. More diagnostic is the SWIR response: bare limestone and granite have high reflectance in Sentinel-2 bands 11 and 12 (1610 nm and 2190 nm), which separates them from both green vegetation and dark shadow. Shadow is itself a problem. Steep north-facing cliffs in the northern hemisphere may be in shadow for most of the day during winter acquisitions, and the 10 m pixel cannot resolve narrow ledges at all.
A practical workflow stacks seasonal Sentinel-2 composites (summer maximum NDVI, winter minimum NDVI, annual SWIR median) and combines them with the DEM-derived slope and aspect layers. The output is a per-pixel probability surface for bare exposed rock on terrain steep enough to exclude most predators. This is the range-scale suitability map. It is not a nest-location map.
The resolution gap: where commercial imagery earns its cost
The transition from candidate-cliff to confirmed nest site requires imagery that can resolve a 1 m ledge. Pléiades at 0.5 m panchromatic resolution can detect nest-platform features and, in favourable conditions, distinguish a large stick nest or whitewash (uric acid deposits from roosting) from bare rock. SPOT 6/7 at 1.5 m is marginal for nest detection but adequate for confirming ledge width and aspect at individual cliff sections.
Commercial tasking is expensive relative to open-data analysis, which is why the two-stage workflow matters economically. Range-scale DEM and Sentinel-2 analysis reduces the area requiring commercial tasking by an order of magnitude or more. A mountain range of 5,000 km² might yield 200 candidate cliff sections from the open-data stage; commercial tasking of those sections at high resolution is a tractable and costed exercise. Attempting to cover the full range at Pléiades resolution without pre-screening is neither practical nor affordable for most conservation budgets.
Timing matters for commercial tasking. Acquisitions in early spring, before deciduous vegetation leafs out, maximise rock-face visibility. Mid-morning local time minimises shadow on south-facing cliffs. These constraints should be written into any tasking specification.
Honest limits of the combined approach
Cloud cover is a persistent problem in montane environments. Sentinel-2's 5-day revisit degrades to weeks of usable acquisitions in persistently cloudy ranges such as the Himalayas or parts of the Andes. Multi-year compositing helps but introduces phenological mixing. SAR-derived DEMs are cloud-independent, which is why TanDEM-X remains the elevation backbone even where optical data is compromised.
The spectral approach cannot distinguish occupied from unoccupied cliffs. A cliff that scores highly on every suitability metric may be vacant due to historical disturbance, food availability, or factors invisible to any satellite. The suitability map predicts physical capacity, not occupancy. Occupancy monitoring requires either field survey or very-high-resolution repeated tasking at known sites to detect nest construction activity.
Satellize's analytics pipeline can run the open-data stages of this workflow, including DEM processing, spectral index compositing, and suitability scoring, across entire mountain ranges, and can coordinate commercial tasking for shortlisted sites on client licence. The workflow is structurally similar to the spatial covariate extraction used in the Tonga crop-estimation programme, adapted here for terrain geometry rather than agricultural phenology.
From suitability map to conservation decision
The output of the full pipeline is a ranked GIS layer of candidate cliff sections, each scored on slope, aspect, bare-rock fraction, ruggedness, and distance to known disturbance sources such as roads or infrastructure. Conservation managers can filter by species-specific thresholds: Bearded Vultures, for instance, show strong preference for cliffs above 1,500 m with southerly aspect in European populations, while Peregrine Falcons tolerate a much wider range of cliff heights and aspects.
The ranked layer feeds directly into survey prioritisation. Field teams visit the highest-scoring unconfirmed sites first, which is both more efficient and more defensible to funders than ad hoc survey design. Sites confirmed as occupied can then be entered into a monitoring programme using repeated commercial tasking to track nest activity across breeding seasons. Sites with high suitability scores but no current occupancy are candidates for reintroduction assessment or disturbance-reduction intervention.
Typical figures
| Elevation product spatial resolution | 12 m (TanDEM-X WorldDEM); 30 m (Copernicus DEM GLO-30, free) |
| Optical mapping resolution | 10 m (Sentinel-2 visible/NIR); 20 m (Sentinel-2 SWIR/red-edge) |
| Site-confirmation resolution | 0.5 m panchromatic (Pléiades); 1.5 m (SPOT 6/7) |
| Sentinel-2 revisit | 5 days at equator with both satellites; effective cloud-free revisit varies by region |
| Minimum detectable cliff band (open data) | Approximately 60 m width at 30 m DEM posting; approximately 25 m at 12 m posting |
| Minimum resolvable ledge feature | ~0.5 m (Pléiades); sub-metre ledges invisible in all freely available products |
| Spectral bands used for bare-rock index | Sentinel-2 B4 (665 nm), B8 (842 nm), B11 (1610 nm), B12 (2190 nm) |
| DEM vertical accuracy (TanDEM-X WorldDEM) | Absolute: ~2 m LE90; relative: ~1 m LE90 over flat terrain; degraded on steep slopes |
| Archive depth (Sentinel-2) | From 2015 (Sentinel-2A); 2017 for dual-satellite coverage |
| Deliverable formats | GeoTIFF suitability rasters, GeoPackage ranked cliff-section vectors, PDF site reports |
Analytics Satellize can run
| Range-scale cliff suitability map | DEM slope/aspect extraction plus Terrain Ruggedness Index; thresholded by species-specific published habitat parameters | GeoTIFF probability raster and ranked vector polygon layer of candidate cliff sections |
| Bare-rock fraction index | Multi-season Sentinel-2 NDVI and SWIR composite; spectral unmixing or threshold classification to separate rock, vegetation and shadow | Per-cliff-section bare-rock percentage attribute in GIS vector layer |
| Aspect and solar-radiation model | DEM-derived aspect raster combined with latitude-corrected solar irradiance modelling to score thermal suitability by cliff face | Aspect-weighted suitability score appended to candidate cliff layer |
| Commercial tasking shortlist and acquisition specification | Ranked filtering of open-data suitability layer to identify highest-priority sites for sub-metre confirmation; tasking parameters (time of day, off-nadir angle, season) derived from shadow-avoidance geometry | Tasking specification document and prioritised site list with coordinates |
| Site-level nest-platform confirmation | Object-based image analysis of Pléiades or SPOT imagery to identify ledge width, whitewash signatures and nest-structure features | Per-site confirmation report with annotated imagery and nest-probability score |
| Disturbance-proximity scoring | Buffer analysis against publicly available road, infrastructure and settlement layers; combined with suitability score to flag high-quality but high-risk sites | Risk-weighted suitability attribute in final GIS layer |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.